predict.int_conformal_full: Prediction intervals from conformal methods

In probably: Tools for Post-Processing Class Probability Estimates

Prediction intervals from conformal methods

Description

Prediction intervals from conformal methods

Usage

## S3 method for class 'int_conformal_full'
predict(object, new_data, level = 0.95, ...)

## S3 method for class 'int_conformal_cv'
predict(object, new_data, level = 0.95, ...)

## S3 method for class 'int_conformal_quantile'
predict(object, new_data, ...)

## S3 method for class 'int_conformal_split'
predict(object, new_data, level = 0.95, ...)

Arguments

object	An object produced by predict.int_conformal_full().
new_data	A data frame of predictors.
level	The confidence level for the intervals.
...	Not currently used.

Details

For the CV+. estimator produced by int_conformal_cv(), the intervals are centered around the mean of the predictions produced by the resample-specific model. For example, with 10-fold cross-validation, .pred is the average of the predictions from the 10 models produced by each fold. This may differ from the prediction generated from a model fit that was trained on the entire training set, especially if the training sets are small.

Value

A tibble with columns .pred_lower and .pred_upper. If the computations for the prediction bound fail, a missing value is used. For objects produced by int_conformal_cv(), an additional .pred column is also returned (see Details below).

See Also

int_conformal_full(), int_conformal_cv()

probably documentation built on July 10, 2023, 2:03 a.m.